We should generalise the cleaning of measurements to remove the measurement/probe overhead from all results in a systematic manner. The best way would be to have an empty test run for the capturing with the current configuration for the metrics (as different metrics enabled causes different impact in the probe effect).
We currently do this for sys calls, but should generalise it for all metrics.
We should generalise the cleaning of measurements to remove the measurement/probe overhead from all results in a systematic manner. The best way would be to have an empty test run for the capturing with the current configuration for the metrics (as different metrics enabled causes different impact in the probe effect).
We currently do this for sys calls, but should generalise it for all metrics.